HAL Id: inria-00563762
https://hal.inria.fr/inria-00563762
Submitted on 8 Feb 2011
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Stratégie de Placement des Puits Mobiles dans les
Réseaux de Capteurs sans Fil pour Bâtiments
Leila Ben Saad, Bernard Tourancheau
To cite this version:
Leila Ben Saad, Bernard Tourancheau. Stratégie de Placement des Puits Mobiles dans les Réseaux de
Capteurs sans Fil pour Bâtiments. JDIR, 2010, Sophia-Antipolois, France. �inria-00563762�
Strat´egie de Placement des Puits Mobiles dans les
R´eseaux de Capteurs sans Fil pour Bˆatiments
Leila Ben Saad, Bernard Tourancheau
LIP UMR 5668 of CNRS-ENS Lyon-INRIA-UCB LyonLyon, France
{Leila.Ben.Saad, Bernard.Tourancheau}@ens-lyon.fr
Abstract—Le besoin des r´eseaux de capteurs sans fil croit tr`es
rapidement dans un large ´eventail d’applications industrielles. Parmi celles-ci se trouve l’observation, le suivi des donn´ees physiques et l’automatisation des bˆatiments. Dans ces r´eseaux, un grand nombre de capteurs transmettent via multi sauts les donn´ees collect´ees vers le puits le plus proche. Les capteurs qui sont proches des puits ´epuisent leurs r´eserves d’´energie beaucoup plus rapidement que les capteurs distants car ils ont une charge de trafic tr`es importante. Ceci est d ˆu au fait qu’ils transmettent leurs propres donn´ees ainsi que les donn´ees des capteurs ´eloign´es provoquant ainsi pr´ematur´ement la fin de la dur´ee de vie du r´eseau. Le d´eplacement p´eriodique des puits permet de r´esoudre ce probl`eme en distribuant la charge du trafic entre les capteurs et am´eliorer ainsi la dur´ee de vie du r´eseau. Dans ce travail, nous proposons un nouvel algorithme qui d´etermine le positionnement de plusieurs puits mobiles dans un r´eseau large ´echelle afin d’augmenter la dur´ee de vie du r´eseau. Son principe se base sur le d´eplacement r´egulier des puits vers les capteurs distants qui ont le plus grand nombre de sauts `a faire pour atteindre le puits le plus proche. Nous avons ´evalu´e les performances de notre solution par des simulations et compar´e avec d’autres strat´egies. Les r´esultats montrent que notre solution am´eliore consid´erablement la dur´ee de vie du r´eseau et ´equilibre notablement la consommation d’´energie entre les nœuds. Ces r´esultats sont tr`es utiles pour le d´eploiement r´eel de r´eseaux de capteurs sans fil au sein des bˆatiments.
Index Terms—R´eseaux de capteurs sans fil, positionnement des
puits, puits mobiles, dur´ee de vie du r´eseau.
I. INTRODUCTION
Wireless Sensor Networks (WSNs) deployment inside build-ings is a very challenging problem. In fact, such networks are formed by a large number of tiny battery-operated sensors which have a limited and non-renewable energy supply be-cause it is usually impractical and even impossible to replace or recharge their batteries. The sensors near the sink are more likely to use up their energy much faster than distant nodes because they carry heavier workloads due to forwarding the data of nodes farther away as well as their own data. Therefore, they become hotspots. The hotspot problem prevents farther nodes to relay their data to the sinks. Consequently, the network lifetime ends prematurely.
More and more efforts have been done recently to improve the lifetime of WSNs. Many communications protocols have been proposed including among others topology control[1][2], routing[3][4] and clustering[5] etc. However, further improve-ment can be achieved if we relocate the sinks in order to change over time the nodes located close to them.
In order to solve the hotspot problem, many researchers propose to place more sensor nodes around the sink[6][7]. However, these solutions are not always feasible in practice and result in unbalanced sensing coverage over different regions of the network. Another proposed solution is to make multiple fixed sinks cooperate with each other to lighten the load and distribute it among the nodes[8][9].
Most of published works suggest to move a single sink to improve the network lifetime[10][11][12][13][14]. But, very few studies focused on the mobility of multiple sinks. Some proposed algorithms find the locations of mobile sinks by solving a mathematical model[15][16]. In [15], the algorithm minimizes the average distances between sensors and closest sinks. In [16], the algorithm selects the locations of sinks in the periphery of the network in such way the difference between the maximum and the minimum residual energy of nodes is minimized. To find the optimal placement of mobile sinks, some researches formulate the problem as an Integer Linear Program ILP[17][18][19] or Linear Program LP[20]. However, the drawback of the LP and ILP based solutions is that they are hard to compute in networks with thousands of sensors due to their high resolution complexity.
Some other works make a moving decision according to the complete knowledge of the energy distribution of the sensors. In [21], the sinks move towards the nodes that have the highest residual energy. But, this strategy requires that the sensors send periodically to the sink additional information about their energy level to allow the sink to found out the nodes which have the highest energy. By doing so, a lot of energy will be wasted.
In this work, our purpose is to determine where to place multiple sinks inside buildings, how long they have to stay in certain locations and where to move them to extend the network lifetime. To answer these questions, we propose a new scalable multi-sink heuristic algorithm (Hop) which regularly moves the sinks towards the distant nodes which have lower load contrary to nodes near the sinks. This approach scales to thousands of nodes and prevents from sending additional information about the energy level of each sensor which is what was done in previous solutions. Simulation results demonstrate that by relocating the mobile sinks according to our proposed algorithm, the energy consumption is balanced among the sensors which lead to significant increase of the network lifetime.
The rest of this paper is organized as follows. In Section 2, we describe the network model including the major assump-tions. Section 3 presents our proposed multi-sink heuristic algorithm. Section 4 evaluates the performance of the pro-posed approach and presents the simulation results. Section 5 concludes the paper.
II. NETWORKMODEL
In order to deploy sensors and sinks inside buildings, we made the following assumptions for the network model.
- We assume N sensors statically placed in a bi-dimensional grid (LxL) of same size cells constructed from the building plan as shown in the Figure 1.
Fig. 1. 10x10 grid of cells with sensors in office building (L=10)
- All sensors have a limited initial energy e0(J) and a fixed
transmission range r (m) equal to the distance between two nodes (i.e, cell size).
- A time-driven application is considered where each sen-sor regularly generates the same amount of data gr(bit/s).
- M sinks keep moving in the grid from one node to another one until the network lifetime end.
- The network lifetime is defined as the time until the first sensor dies (i.e, it uses up its residual energy).
- The sinks should stay at a certain location for at least a duration of time T (sojourn time). At the end of this duration, they can change their locations.
- The traveling time of sinks between sensor nodes is considered negligible for analytical simplicity.
- The sensor nodes which are not co-located with any sinks inside the grid, relay their generated data via multiple hops to reach the nearest sink using the shortest path routing protocol.
- In our routing protocol, we consider only the two paths along the perimeter of the rectangle, i.e., paths 1 and 2 in Figure 2. These two routes are considered equivalent. - An ideal MAC layer with no collisions and
retransmis-sions is assumed.
- Only the energy consumption for communication is considered. Let eT (J/bit) be the energy consumption
coefficient for transmitting one bit and eR (J/bit) be the
energy consumption coefficient for receiving one bit.
Fig. 2. Path selection
III. MULTI-SINKHEURISTICALGORITHM(HOP) The purpose of the multi-sink heuristic algorithm is to find the best way to move the sinks in order to improve the lifetime of large scale sensor networks. Our approach is based on number of hops and consists in relocating periodically the sinks towards the distant nodes. The difference between our strategy and what was already proposed is that there is no need for the sensors to drain their energy in sending additional information about their energy level. Each sink knows its own position, others sinks positions and the locations of all the sensors. Therefore, from the number of hops to reach the nearest sink, it is possible to guess which sensors are distant and may have more residual energy.
The algorithm begins with an initialization phase where the sinks are placed at their optimal locations in terms of hop counts. Then, for each sensor, the number of hops to reach the nearest sink is computed. Next, the nodes are sorted with decreasing number of hops in order to determine the distant nodes from the sinks. Afterwards, the first sink will be relo-cated at the farthest node. The second sink will be relorelo-cated at the following distant node but respecting the condition that the number of hops between the two new locations must be upper than minimum number of hops minhop. The third sink
is relocated with the same manner in such way the distance between the three new positions of sinks is upper than minhop.
The remaining sinks are relocated with the same way. All the chosen positions of sinks at each period are saved in a list. In the case that the selected positions of sinks where already chosen in previous periods, the algorithm chooses as the first sink location a node which has not been chosen before and determines the locations of the other sinks. If all sensor nodes where already visited by the sinks, the chosen list is emptied. The same operations are repeated at the beginning of each new period T.
The pseudocode of the algorithm is shown in the Figure 3. The algorithm is implemented in a distributed manner. In the sense that it does not determine the sinks locations based on the knowledge of the global network parameters. But, each sink has to know only the initial starting locations of sensors and sinks. Throughout the network lifetime, each sink is able to compute autonomously its next position and move there.
To have a better idea about the algorithm, an example of a network with 100 sensors and 3 mobile sinks is provided. The sinks locations pattern obtained by the execution of our
Multi-Sink Heuristic Algorithm (Hop)
1: place the M sinks at optimal starting locations in LxL grid 2: add starting locations of the sinks to chosenlist
3: minhop=
½
L − 1, M = 2
[2(L − 1)/(M − 1)], M >= 3 4: while new period T do
5: for i = 1 to N do
6: nearestK = nearest sink for sensor nodei
7: nhop = number of hops between nodei and nearestK
8: add (nodei, nhop) to nodelist
9: end for
10: sort nodes of nodelist in decreasing order of nhop
11: p1= select first node of nodelist in the sorted order
12: add p1 to selectlist
13: for i = 2 to M do
14: pi = select next node from nodelist having minimum
number of hops with nodes in selectlist >= minhop
15: add pi to selectlist
16: end for
17: if nodes from selectlist not in chosenlist then
18: add nodes from selectlist to chosenlist
19: else
20: empty selectlist
21: p1= select not chosen node from nodelist
22: go to line 12 23: end if
24: if chosenlist contains all nodes then
25: empty chosenlist
26: end if
27: for i = 1 to M do
28: move sinki to node pi from selectlist
29: end for 30: end while
Fig. 3. Hop algorithm
Fig. 4. The sinks locations pattern in 10x10 grid network (L = 10)
algorithm during the three first periods T are represented in the Figure 4. The sinks were initially placed at their optimal locations in sensor nodes 15, 45 and 86. The minimum number
of hops between the locations of sinks minhop was fixed to
9.
IV. SIMULATION RESULTS
To analyze the performances of our proposed algorithm, we built a simulator in Java environment with variable number of sensors and sinks deployed on different grid sizes.
To compute the energy consumption, we used the same model described in [19]. We chose realistic parameter assump-tions for the network. The initial energy at each node is equal to energy found in two Alkaline batteries AA of 1.5V usually 2600mAh i.e., eo= 28080 J. The energy consumption
coeffi-cient for transmitting and receiving one bit was chosen the same as vendor-specified values for the Chipcon CC2420[22] where eT = 0.225 10−6J/bit and eR= 0.2625 10−6J/bit. We
fixed the minimum duration of sojourn time T of the sinks to 30 days because it is economically not easy for technicians to relocate sinks in buildings very often. The sensor transmission range is r = 10 m. The rate at which data packets are generated is gr= 1 bit/s. Notice that real micro-controllers stop
running when the battery voltage is below a given threshold. This depends on the micro-controllers and can not be taken into account here.
To show the efficiency of our proposed scheme, we evalu-ated its performance by making a comparative study with four other schemes.
Thus, the following schemes were implemented : 1) Static: Static sinks placed in optimal locations[19] 2) Periphery: Sinks moving in the periphery of the network 3) Random: Sinks moving randomly
4) Hop: Sinks moving according to our algorithm 5) Opt: Sinks moving according to ILP solution[19] In the following sections, the network lifetime, the energy consumption and the residual energy at each sensor are inves-tigated.
A. The network lifetime
We evaluated the network lifetime of the five schemes on networks of 100 sensors with increasing number of mobile sinks.
Fig. 5. Network lifetime with 100 sensors (10x10 grid)
Figure 5 shows that the Opt scheme performs better than the others schemes because it selects the optimal locations of
Fig. 6. Network lifetime with increasing number of sinks
sinks and their optimal sojourn times. But, the problem with Opt is that it is based on Integer Linear Program which can handle only small scale networks. All the remaining schemes are scalable but the Hop scheme is the best of them because it achieves longer lifetime. The lifetime improvements achieved by Hop in 10x10 grid network with 3 mobile sinks are almost 38 % against Random, 64 % against Periphery and almost 170 % against Static. The gap between Hop and Opt in network lifetime is about 18 %.
We investigated the network lifetime of Static, Random, Periphery and Hop schemes in a large scale network with 2500 sensors on 50x50 grid. We varied the number of sinks from 2 to 10.
Figure 6 shows that all schemes improve the network lifetime when the number of sinks increases. Because, using more sinks reduces the average path length between the sensors and sinks and enables to achieve less traffic load to the nodes which extends the network lifetime. Nevertheless, the Hop scheme leads to longer network lifetime than all other schemes. In a network with 2500 sensors and 10 sinks, Hop achieves 29 % of lifetime improvement than when 10 sinks are moving randomly, 42 % than when 10 sinks are moving on the periphery, and 1014 % than when 10 sinks are static.
Fig. 7. Network lifetime with different sinks sojourn times
We studied the network lifetime with different sinks sojourn times (see Figure 7). The number of sinks was fixed to 10 and the number of sensors to 2500. The results indicate that when the the sinks sojourn time increases the network lifetime decreases slowly. In fact, the longer the sojourn time is, the less the sinks movements are which lead to shorter lifetime.
We varied the network size from 10x10 to 50x50 sensors.
Fig. 8. Network lifetime with increasing network size
The number of sinks was fixed to 10 and the sojourn time T to 30 days. We noticed as shown in the Figure 8 that the network lifetime decreases considerably when the network size increases. In fact, the sensors near the sinks must retransmit a higher number of packets from their higher number of neighbors which leads to higher energy consumption.
B. The energy distribution
We analyzed the impact of the five schemes on the energy consumption at lifetime end in a network with 3 mobile sinks and 100 sensors. The distribution of energy consumption when the first sensor dies is depicted in Figures 9, 10, 11, 12 and 13. A light color means a higher percentage of energy consumption.
It is remarkable in all the figures that the energy consump-tion is highly variable and depends on the sinks locaconsump-tions. We notice that the nodes closest the sinks locations have relatively higher energy consumption compared to most of the others because they have to receive and relay all other neighbors data in addition to their own data. This leads them to consume more energy.
Fig. 9. Energy consumption with Static scheme
In Figure 9, we observe that higher percentage of energy consumption is concentrated around three nodes in the grid which are the locations of the sinks whereas the others sensors have a lower amount of energy consumption (dark color).
When the sinks move on the periphery of the network the highest energy consumption occurs in nodes closest to the boundary of the network while the others nodes specially in the center consume less energy as seen in the Figure 10.
Fig. 11. Energy consumption with Random scheme
Figure 11 shows that the energy consumption of Random scheme is more balanced among the nodes than Periphery and Static schemes.
Fig. 12. Energy consumption with Hop scheme
Figure 12 shows that Hop scheme results in a better balanc-ing of energy consumption than Periphery, Static and Random schemes. In fact, we notice a larger area with light color.
Fig. 13. Energy consumption with Opt scheme
Opt scheme balances almost perfectly the energy consump-tion among the nodes. In fact, the majority of the nodes deplete their energy at the same time except the four nodes in the corners as shown in the Figure 13. However, this scheme is restricted only to small scale networks.
The distribution of the residual energy at each sensor node in 10x10 grid with 3 mobile sinks was also studied (see Figures 14, 15, 16, 17 and 18).
The results show that the percentages of the residual energy that remained unused at the network lifetime end are 71 %,
Fig. 14. Residual energy with Static scheme
Fig. 15. Residual energy with Periphery scheme
Fig. 16. Residual energy with Random scheme
Fig. 17. Residual energy with Hop scheme
Fig. 18. Residual energy with Opt scheme
45 %, 31 %, 17 % and 3 % respectively for Static, Periphery, Random, Hop and Opt schemes. Moreover, the numbers of sensors which have more than 50 % of their initial energy at network lifetime end are 80, 36, 20, 4 and 4 respectively for Static, Periphery, Random, Hop and Opt schemes.
The two schemes Opt and Hop result in a better distribution of residual energy compared to Periphery, Random and Static. Nevertheless, Opt can be efficient only for small networks
contrary of Hop which can scale to thousands of sensors. We analyzed the impact of the scalable schemes Static, Periphery, Random and Hop on the energy consumption at lifetime end in a large scale network with 4 mobile sinks and 2500 sensors. The results are similar to what was found in small network with 100 sensors. Hop balances better the energy consumption among the nodes than the other schemes because it leads to a higher number of sensors with very small amount of energy left unused (see the Figure 19).
(a) Static (b) Periphery
(c) Random (d) Hop Fig. 19. Energy consumption in 50x50 grid network
V. CONCLUSION
In this paper, we have proposed an efficient solution to extend the lifetime of large scale WSNs. The proposed heuris-tic algorithm regularly moves the sinks towards the distant nodes. This approach, based on number of hops, prevents from sending additional information about the energy level of each sensor. We evaluated the performance of our algorithm by simulation in a network with thousands of sensors and compared it with others schemes: Static, Periphery, Random and Opt. The results show that it extends significantly the lifetime of the network and balances notably the energy consumption among the nodes. Our solution is very simple and useful for wireless sensor network deployment inside buildings because it is scalable and achieves 1014 % lifetime improvement when we deploy 10 mobile sinks instead of 10 static ones in a network with 2500 sensors.
In our future work, we intend to study the management of multiple sinks moving according to our proposed solution in a 6lowpan-based wireless sensor networks for buildings. We would like to focus on how IPv6 mechanisms could support efficiently sinks periodic mobility.
REFERENCES
[1] J. Pan, Y. T. Hou, L. Cai, Y. Shi, and S. X. Shen, “Topology control for wireless sensor networks,” in Proceedings of the 9th annual international
conference on Mobile computing and networking MobiCom, 2003, pp.
286–299.
[2] X. yang Li, W. zhan Song, and Y. Wang, “Topology control in hetero-geneous wireless networks: Problems and solutions,” in Proceedings of
the 23 rd Joint Conference of the IEEE Computer and Communications Societies INFOCOM, 2004.
[3] A. Sankar and Z. Liu, “Maximum lifetime routing in wireless ad-hoc networks,” in Proceedings of the 23 rd Annual Joint Conference of the
IEEE Computer and Communications Societies INFOCOM, vol. 2, 2004,
pp. 1089–1097.
[4] K. Fodor and A. Vid´acs, “Efficient routing to mobile sinks in wireless sensor networks,” in WICON ’07: Proceedings of the 3rd international
conference on Wireless internet. ICST, Brussels, Belgium, Belgium: ICST (Institute for Computer Sciences, Social-Informatics and Telecom-munications Engineering), 2007, pp. 1–7.
[5] O. Younis and S. Fahmy, “Distributed clustering in ad-hoc sensor networks: A hybrid, energy-efficient approach,” in Proceedings of the 23
rd Annual Joint Conference of the IEEE Computer and Communications Societies INFOCOM, 2004, pp. 629–640.
[6] X. Wu, G. Chen, and S. K. Das, “On the energy hole problem of nonuniform node distribution in wireless sensor networks,” IEEE
International Conference on Mobile Adhoc and Sensor Systems MASS,
2006.
[7] ——, “Avoiding energy holes in wireless sensor networks with nonuni-form node distribution,” IEEE Trans. Parallel Distrib. Syst., 2008. [8] H. Kim, Y. Seok, N. Choi, Y. Choi, and T. Kwon, “Optimal multi-sink
positioning and energy-efficient routing in wireless sensor networks,”
Lecture Notes in Computer Science LNCS, 2005.
[9] E. Oyman and C. Ersoy, “Multiple sink network design problem in large scale wireless sensor networks,” IEEE International Conference
on Communications, vol. 6, pp. 20–24, 2004.
[10] B. Wang, D. Xie, C. Chen, J. Ma, and S. Cheng, “Employing mobile sink in event-driven wireless sensor networks,” IEEE Vehicular Technology
Conference VTC, 2008.
[11] Z. M. Wang, S. Basagni, E. Melachrinoudis, and C. Petrioli, “Exploiting sink mobility for maximizing sensor networks lifetime,” Proceedings of
the 38th Annual Hawaii International Conference on System Sciences HICSS, 2005.
[12] K. Akkaya and M. Younis, “Sink repositioning for enhanced per-formance in wireless sensor networks,” Elsevier Computer Networks
Journal, 2005.
[13] S. Basagni, A. Carosi, E. Melachrinoudis, C. Petrioli, and Z. M. Wang, “Controlled sink mobility for prolonging wireless sensor networks lifetime,” Wireless Networks, 2007.
[14] Y. Bi, J. Niu, L. Sun, W. Huangfu, and Y. Sun, “Moving schemes for mobile sinks in wireless sensor networks,” Performance, Computing, and
Communications Conference, 2007.
[15] Z. Vincze, R. Vida, and A. Vidacs, “Deploying multiple sinks in multi-hop wireless sensor networks,” IEEE International Conference on
Pervasive Services, 2007.
[16] A. Azad and A. Chockalingam, “Mobile base stations placement and energy aware routing in wireless sensor networks,” IEEE Wireless
Communications and Networking Conference WCNC, 2006.
[17] S. R. Gandham, M. Dawande, R. Prakash, and S. Venkatesan, “Energy efficient schemes for wireless sensor networks with multiple mobile base stations,” The IEEE Global Telecommunications Conference
GLOBE-COM, vol. 1, 2003.
[18] W. Alsalih, S. Akl, and H. Hassanein, “Placement of multiple mobile base stations in wireless sensor networks,” IEEE International
Sympo-sium on Signal Processing and Information Technology, 2007.
[19] L. Ben Saad and B. Tourancheau, “Multiple mobile sinks positioning in wireless sensor networks for buildings,” in The Third International
Conference on Sensor Technologies and Applications SensorComm,
2009.
[20] S. Basagni, A. Carosi, C. Petrioli, and C. Phillips, “Moving multiple sinks through wireless sensor networks for lifetime maximization,” in 5th
IEEE International Conference on Mobile Ad Hoc and Sensor Systems MASS, 2008.
[21] M. Marta and M. Cardei, “Improved sensor network lifetime with multiple mobile sinks,” Pervasive and Mobile computing, 2009. [22] Chipcon. CC2420 2.4 GHZ IEEE 802.15.4 / ZigBee-ready RF